Appearance
Agentkit
Agentkit is a lightweight, model-agnostic framework for building AI agents in Python. The language model decides what tools to call, in what order, until it reaches a final answer — no brittle chains or rigid pipelines needed.
What you can build
| Pattern | Description |
|---|---|
| Conversational assistants | Single-agent chatbots that call tools |
| Autonomous task agents | Multi-step reasoning with memory and state |
| Multi-agent systems | Graphs of agents that collaborate or compete |
Key features
| Feature | Description |
|---|---|
| Model Agnostic | Amazon Bedrock, Anthropic, OpenAI, Gemini, Ollama, LiteLLM, and more |
| Tool system | Python decorators, directory hot-reload, MCP server support |
| Multi-agent | Graph (deterministic) and Swarm (autonomous) patterns |
| Streaming | Async token streaming |
| Hooks | Strongly-typed lifecycle events for observability and control |
| Sessions | Built-in conversation persistence with file, S3, or custom backends |
| Checkpoints | Thin ReAct-loop bookmarks for mid-tool crash recovery under a session |
| Permissions | Declarative allow / deny / interrupt rules for host files and commands |
| Observability | OpenTelemetry tracing and metrics out of the box |
Minimal example
Default model is Amazon Bedrock
Agent() with no model= uses Amazon Bedrock (Claude Sonnet 4.6 in us-west-2). Configure AWS credentials (aws configure or AWS_* env vars), or pass an explicit model — see Installation and Model Providers.
python
from elsai import Agent
agent = Agent()
agent("Explain quantum entanglement in one paragraph")Every additional capability — tools, memory, multi-agent orchestration — layers on top without rewriting existing code.
Installation
bash
pip install --extra-index-url https://elsai-agents.elsai.ai/root/ elsai-agents==0.3.1
pip install --extra-index-url https://core-packages.elsai.ai/root/ "elsai-model[bedrock]==2.1.0"For other providers, install the matching extra (for example "elsai-model[openai]==2.1.0"). See Installation.
Next steps
- Quickstart — build your first agent in 5 minutes
- Agent concepts — understand the agent loop
- Tools — add capabilities to your agents
- Multi-Agent — orchestrate multiple agents
- Checkpoints — mid-tool crash recovery bookmarks